About Karaoke-style movie generation

[ashyrv]

Hey everyone! I have a question about the karaoke-style speech recognition. Do you think it will work in real time? It seems it's really good with recorded audio but will it work in real-time speech? Any feedback or answers appreciated.

Thank you in advance!

[ulatekh]

Whisper is not real-time, nor is it likely to ever be real-time. It works by processing audio 30 seconds at a time, and that processing, even when GPU-accelerated, can take a significant fraction of a second, or even several seconds.

[ashyrv]

Thanks

…

On Wed, 5 Jun 2024, 00:05 ulatekh, ***@***.***> wrote: Whisper is not real-time, nor is it likely to ever be real-time. It works by processing audio 30 seconds at a time, and that processing, even when GPU-accelerated, can take a significant fraction of a second, or even several seconds. — Reply to this email directly, view it on GitHub <#1206 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AUDBSE67MG25ECR2KBXS4RTZFYT2ZAVCNFSM6AAAAAA36EDHT6VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4TMNRZHA2TM> . You are receiving this because you authored the thread.Message ID: ***@***.***>